What AI Certifications Engineers Can Take in 2026

🌏 中文版

All three major cloud vendors overhauled their AI certification tracks in the first half of 2026, so what you can actually register for looks quite different from a year ago. This piece lists every exam an engineer can sit right now, with prices, validity, and registration gates taken from official announcements.

Two things worth knowing before you pay: Google's PMLE exam guide has renamed Vertex AI throughout, which makes pre-mid-2026 study material worthless, and the iPAS intermediate certificate is valid 5 years, not permanently as commonly written.

The Table: What You Can Register For

Microsoft and Anthropic are absent from this table: Microsoft's track turned over entirely this year and the codes need handling separately, and Anthropic's four are open only to partner organizations; each has its own section below. iPAS is Taiwan-specific and follows different rules, so it gets its own section too. The NVIDIA row lists only the entry-level exam — the catalog actually holds twelve, tabled further down. Snowflake, Oracle, IBM, Salesforce, and GitHub are in "Other Ecosystems," and the governance line (IAPP, ISACA, CertNexus) is a different kind of credential entirely, so it gets its own section.

Which One to Take

Confirm which cloud your company actually runs first, then pick the certification. Reversing that order leaves you with an AWS credential that carries almost no negotiating power at an Azure shop.

If you want foundations rather than a credential, the site also has which AI courses to take in 2026 (in Mandarin), which reorders the official OpenAI, Anthropic, and Google offerings by capability tier.

This table answers "which one", not "how to prepare". For that the site has a whole series — AI Certification Prep: start with how to set the exam date and study order, then read the path for your exam, one per certification and built on each vendor's official exam-guide domain weights, plus five posts comparing what the exams share. If you are undecided, the per-vendor chooser is the fastest way in:

• Which of the three AWS certifications — including how the MLA-C02 beta differs from the GA exam
• Which of the four Microsoft certifications — where the code-first and low-code lines diverge
• Which of the four NVIDIA certifications — two professionals are not registrable yet and their official training is paid; the two associates now recommend one free open-source set of materials plus one course
• Which of the four Claude certifications — individuals cannot register; Claude Partner Network organizations only

Technical topics tested across several exams are pulled into five deep dives: multi-agent architecture, RAG and retrieval evaluation, AI governance frameworks, prompt and context engineering, and cost and latency optimization.

AWS: Three Exams Separated by Price and Expected Experience

AIF-C01 is the entry point at $100, 90 minutes, 65 questions. MLA-C02 moves up to ML engineering practice: $150 and 65 questions at GA, with an English beta open now ($75, 170 minutes, 85 questions). At the top sits AIP-C01 at $300, 180 minutes, 75 questions, where AWS expects 2 years of AWS experience plus 1 year of GenAI development — the only one of the three aimed squarely at GenAI application development.

MLA is now the C02 English beta. MLA-C01 in English ended on September 28, 2026, and the official certification page has delivered the MLA-C02 beta (exam code ME1-C02) since September 29; Japanese, Korean, and Simplified Chinese stay on C01 until C02 reaches general availability. All four C02 domain names now mention AI or foundation models; what it tests and how to prepare are in the MLA-C02 preparation path.

AIF-C01 is offered in Traditional Chinese. It runs in 12 languages including Traditional Chinese — the only one of AWS's three that does; MLA and AIP-C01 have no Traditional Chinese version; Google's PMLE is English and Japanese only. Sitting an exam in your first language is worth factoring into the time estimate.

A failed attempt means a 14-calendar-day wait and the full fee again. That is the After Testing policy; attempts are unlimited, and results post within five business days.

All three are valid 3 years, but the renewal rules are more layered than most roundups suggest. The official recertification page spells them out:

The AI trio has no "take a course instead" option. AWS as a program offers two paths — renew (+3 years) and maintain (+1 year, completed on AWS Skill Builder with a paid subscription) — but maintain is listed only for SAA, Developer, CloudOps/SysOps, SAP, and DOP. The Options column for AIF-C01, MLA, and AIP-C01 contains exams only.

Passing a higher exam renews the lower ones. This is the cheapest path and the least frequently written down:

One AIP-C01 pass therefore pushes AIF-C01, MLA, and Data Engineer – Associate three years out as well.

Renewals come with a 50% voucher. The Cost column for all three reads "Use the 50% discount voucher in your AWS Certification Account" — so the three-year renewal is $50 / $75 / $150, not list price.

Preparation paths: AIF-C01, MLA-C02, and AIP-C01; the trade-offs are in which of the three AWS certifications.

Google: PMLE Is What You Can Take Now, and the Material Must Postdate Mid-2026

Google has two AI certifications open for registration today: Generative AI Leader (foundational) and Professional ML Engineer (professional). A third, Professional Agentic Architect, covers building, orchestrating, and governing agents, and its general-availability registration opens on November 2, 2026. Until then PMLE is the only Google exam an engineer can take to demonstrate GenAI or agent capability, and its new exam guide already includes agentic content.

Generative AI Leader is not worth $99 for an engineer. Google is blunt about the audience:

This certification is for anyone in any job role, with or without hands-on technical experience.

Too little signal for anyone with a programming background. Exam languages are English, Japanese, Spanish, and Portuguese — no Chinese.

PMLE's preparation path is in the Google PMLE prep guide.

Every Product Name in the Exam Guide Was Renamed

This is the one thing to know before registering for PMLE. Most articles describe it as covering "Vertex AI" — but that term has all but vanished from the current exam guide. Comparing the full official exam guide:

At Cloud Next '26, Google folded Vertex AI into Gemini Enterprise Agent Platform, which officially "brings together the best of Vertex AI with transformational new features, including Agent Studio, Agent-to-Agent Orchestration, Agent Registry, Agent Identity, Agent Gateway, Agent Observability."

Practical impact: every PMLE course, online class, and question bank published before mid-2026 uses the old vocabulary. Exam questions use the new names directly, and anyone who only recognizes legacy Vertex AI terminology will stall on "Agent Platform Feature Store." This is the fastest way to waste $200.

Section weights were also rebalanced:

Several topics won't appear in traditional ML material at all: LLM-as-a-judge evaluation, prompt and context engineering, cost / latency / availability optimization for Gemini applications, and fine-tuning Gemini models using BigQuery.

Factor validity into the cost: the official exam terms state professional certifications are valid two years, with renewal opening 60 days before expiry.

The renewal rules changed in 2026, and two details affect the math. The official renewal page says continuing education on Google Skills can add a year of validity, and the table lists foundational, associate, and professional tiers alike — but the caveat on the same page states that only CDL, ACE, PCA, and PDE currently have that option. PMLE is not among them; Google says it plans "to add the Google Skills renewal option to the other certifications at a later date." So PMLE still means retaking the exam today, though the restriction is clearly temporary.

The other commonly missed detail is the discount: the same page states you receive a 50% renewal discount code when you first certify, retrievable from your CM Connect profile. That halves the $200 renewal — provided you don't let the certification lapse more than 30 days, after which you pay full price for the standard exam.

Two Ways to Spend Less on Google

The Get Certified program lets Google Cloud customers take certification prep training at no cost. If your company is a GCP customer, check internally for a slot before paying $200.

The GEAR program, announced at Cloud Next '26 and running on Google Skills, is hands-on agent training whose new learning paths include Introduction to Agents and Google's Agent Ecosystem and Develop Agents with Agent Development Kit (ADK). It maps directly onto PMLE's new agentic topics, and ADK work converts straight into portfolio material.

Microsoft: AI-103 as the Base, Agent Credentials on Top

The current associate-level exam is AI-103, leading to Azure AI Apps and Agents Developer Associate. The official description reads "designing, developing, and deploying advanced Azure AI solutions using Python and Microsoft Foundry" — the core platform is Microsoft Foundry, so preparation looks nothing like the old Azure AI service lineup.

Microsoft's agent track is more complete than anyone else's. Beyond the already-GA AB-620 (AI Agent Builder Associate), there are two expert-level credentials: Multi-Agent AI Solutions Expert (AI-500, beta, officially positioned around "designing, building, and optimizing scalable, production-ready, multi-agent AI systems") and Agentic AI Business Solutions Architect (AB-100).

AI-500 has a hard prerequisite that most roundups omit. The "Certification prerequisites" section of the official page states:

To become a Microsoft Certified: Multi-Agent AI Solutions Expert, you must earn the Microsoft Certified: Azure AI Apps and Agents Developer Associate certification.

You must hold AI-103 first; the page's field reads Prerequisites: 1 certification. Going straight to AI-500 is not possible, so the real cost is $165 (AI-103) plus $165 (AI-500). Microsoft removed the beta label from the official pages at the end of September 2026. The exam is English-only, and the page still says "Learning paths or modules are not yet available for this certification."

AI-500 deserves separate attention. Among the three major clouds, only Microsoft offers an expert-level certification dedicated to multi-agent system architecture, so it differentiates you a tier above the associate-level AI-103. It is not the only advanced agent credential on the market, though. NVIDIA's NCP-AAI, Oracle's Agentic AI track, and Google's new Professional Agentic Architect (GA registration opens November 2, 2026) aim at the same thing; the difference is whose platform you get locked to.

AB-100 has a prerequisite too. Microsoft now states that besides passing AB-100 you must hold one of the associate certifications on its list (AB-620, AI-103, and AI-300 are among them). Details are in the AB-100 preparation path.

Microsoft renews through a free online renewal assessment, rather than retaking the full exam as AWS and Google require; the four Claude exams also renew through a free assessment if you do it on time. The GitHub Copilot certification (GH-300), which sits under the Microsoft umbrella, does not have that advantage yet: 100 minutes, 2-year validity, proctored by Pearson VUE, covering responsible Copilot use, prompt crafting, agent mode and MCP, content exclusions, and audit logs. GitHub says its certifications are moving to Microsoft's recertification process, which is not live yet. Price depends on the country where the exam is proctored; the country selector on the certification page shows $99 for the United States and $50 for Taiwan. Preparation paths: GH-300 and GH-600.

Preparation paths for all four: AI-103, AI-500, AB-620, AB-100; the trade-offs are in which of the four Microsoft certifications. Three more on the Azure AI development track have their own paths: the entry-level AI-901, AI-200 on backend services and vector databases, and AI-300 on MLOps and GenAIOps.

NVIDIA: Twelve Exams, Not Just NCA-GENL

The table above lists only the entry-level one. The official certification catalog actually lists twelve:

Two are worth calling out. NCP-AAI (Agentic AI) is one of the few vendor credentials built directly around multi-agent interaction, scalability, and ethical safeguards at professional level, and $200 is an easier entry than Microsoft's expert line. NCA-GENM (Generative AI Multimodal) is the only one here aimed specifically at systems spanning text, image, and audio — a better match than NCA-GENL if that's what you build.

One practical warning: NCP-AII, NCP-AIN, NCP-ARI, and NCP-AIO run $400–$500 and test GPU cluster deployment, monitoring, and interconnect. Those are infrastructure-team credentials, not application-engineer ones. Picking the wrong tier means paying four times as much for an exam you can't use.

Preparation paths for the four generative-AI exams: NCA-GENL, NCP-GENL, NCP-AAI, NCA-GENM; the trade-offs are in which of the four NVIDIA certifications.

Other Ecosystems: Snowflake, Oracle, IBM, Salesforce

If your company isn't on any of the clouds above, these are the ones that count:

Oracle changed two things this year that affect how you search for material. Per the official announcement, certification titles and badges no longer carry the calendar year (it's "OCI Architect Associate," not "OCI 2026 Architect Associate"), and while Foundations-level courses and exams stay free, Professional-level courses moved to paid access. Most tutorials you'll find still teach exam codes that no longer match the official pages, so compare names before trusting one.

Claude: Four Certifications, All Specs in the Exam Guides

Anthropic announced on July 23, 2026 that the certification program had expanded to four credentials spanning Associate, Developer, and Architect roles. The same announcement notes that since the March launch, "more than 36,000 consultants have received certification across more than 1,300 organizations."

Engineers want Claude Certified Developer: Foundations, officially described as "for engineers building applications with Claude, and includes training on the Claude API, tool use, and agent development."

The fees and validity periods for all four are in their exam guide PDFs, downloadable from the per-certification pages on Anthropic Partner Academy:

All four pass at 720 on a 100–1,000 scaled score, mix multiple-choice with multiple-response items, are delivered online proctored or at a Pearson test center, and report percent-correct by domain on the score report. CCAR-F has one structural quirk: four scenarios drawn from a bank of six.

The certification pages also publish domain weights. CCAR-F breaks down as Agentic Architecture & Orchestration 27%, Claude Code Configuration & Workflows 20%, Prompt Engineering & Structured Output 20%, Tool Design & MCP Integration 18%, Context Management & Reliability 15%; CCDV-F is dominated by Applications and Integration at 33.1%.

Retake rules are on the Pearson VUE page: 14 days after a first failure, 30 after a second, 90 after a third, and at most 4 attempts per exam in any rolling 12-month period.

The registration gate is stated on the Pearson VUE page:

Certification is open to organizations in the Claude Partner Network and counts toward partner program standing.

There is no official route for individuals to register, so plan on needing access through a partner organization. Exam-prep training is likewise open only to Claude Partner Network members; joining the network is free for a company, so the first step is checking whether yours is in it.

Preparation paths for all four: CCAR-F, CCAR-P, CCDV-F, CCAO-F; the trade-offs are in which of the four Claude certifications.

iPAS Intermediate: Taiwan's Local Option, Valid 5 Years

Per the official exam information (in Mandarin), intermediate Subject 1 (AI Technology Application and Planning) is mandatory, plus one of Subject 2 (Big Data Processing and Analysis) or Subject 3 (Machine Learning Technology and Applications), both requiring ≥70. The fee is a 2026–2027 promotional NT$500 per subject (regular price NT$1,500), so NT$1,000 for both, reverting in 2028.

About 25% of the intermediate exam is Python code questions. Per the iPAS notice on code question weighting (in Mandarin), code interpretation questions were added to Subjects 2 and 3 starting with the second 2025 session, as single-answer items and as question sets; a set hangs several questions on one code snippet, so one misreading costs several answers.

Validity is the detail most sources get wrong; even 104's certification guide (in Mandarin) states the certificate is valid permanently. The 2026 iPAS AI Application Planner examination handbook (in Mandarin) is explicit:

Renewal has one engineer-friendly provision: each year of AI-related work experience offsets 8 hours of training. If AI is your day job, five years of experience covers 40 hours, leaving just 8 hours of coursework.

Sessions are few: per the official registration instructions (in Mandarin), the intermediate exam runs only twice in 2026 (May 23 and November 14), and individual registration for the second session closed on 2026-09-22. Miss it and you wait for the next year's sessions.

The Intermediate Certificate Is Split, and Subject Scores Carry Over

Per the 2026 handbook (in Mandarin), intermediate isn't one generic certificate. Subjects 1 + 2 at 70 or above yields "AI Application Planner (Data Analysis)"; Subjects 1 + 3 yields "AI Application Planner (Machine Learning)." The parenthetical shows up on your résumé, so decide which one you want first.

Two money-saving rules that people miss:

• A passing subject score is held for three calendar years. A subject passed in 2026 counts through 2029-12-31, so the next sitting only needs the other subject.
• Older iPAS credentials grant exemptions. Holders who passed Subject 2 of the former Machine Learning Engineer (beginner) can be exempted from intermediate Subject 3; holders who passed Subject 1 of the former Big Data Analyst (intermediate) can be exempted from either Subject 2 or Subject 3 (one only). Either way, passing Subject 1 at 70 or above is enough to earn the certificate. Exemption eligibility only runs through the end of 2027.

Beginner Level: Permanent, and Worth Banking

The beginner level covers two subjects — Introduction to AI Fundamentals, and Generative AI Applications and Planning — at 75 minutes each with a 70-point pass mark, and runs four times in 2026 (March 21, May 16, August 15, November 7), three times as often as intermediate. The beginner certificate is valid permanently with no renewal, unlike intermediate's 5 years; the validity table above lists both, so don't conflate them.

For a faster local credential there are two non-iPAS options from the Computer Skills Foundation: TQC AI Applications and Technology (Practical AI1 / Advanced AI2 / Professional AI3, NT$1,000 each, 50 knowledge questions) and TQC+ Artificial Intelligence: Machine Learning Python 3 (PML3, NT$1,800, 60 minutes, hands-on coding only). PML3 is the only local exam here that tests implementation, but neither carries the weight of an iPAS certificate issued by the Ministry of Economic Affairs.

Governance, Audit, Security: A Completely Different Line

Everything above asks "can you build it." This section asks "can you sign off on it." Companies adopting AI now need someone accountable for the risk, and these credentials barely overlap with the engineering ones — two of them require you to already hold something else.

What sets CAIP apart isn't the content, it's the accreditation. It is ANAB-accredited under ISO/IEC 17024, the only credential in this entire article that is — a vendor certification derives its authority from the vendor, while CAIP's comes from a third party having audited how it writes and scores questions. Procurement, audit, and compliance contexts care about the latter. The content itself is vendor-neutral end-to-end ML: problem framing 26%, feature engineering 20%, training and tuning 24%, operationalization 30%.

Read the ISACA gates carefully: these are add-ons to credentials you already hold, not entry points. Without CISM / CISSP / CISA you cannot sit them, and the prerequisite itself represents years of experience plus an exam. Conversely, if you already hold one, the marginal cost of adding an AI credential is low and the CPE counts toward both.

AIGP is the most policy- and regulation-oriented of the group, covering responsible development, deployment, and governance frameworks. It suits legal, risk, and PM-plus-governance roles — it is not a way to demonstrate engineering ability.

Beyond the Exam Fee: Course Costs and Ways to Pay Less

Courses are a separate cost from the exam fee, and the two are independent: no vendor requires a course before you register. The free tier is enough in most cases.

The AWS $29 month is one of the few payments worth making. The full practice exam sits in the paid tier, and the 20 free questions are not enough to tell whether you would pass. Subscribe a month before the exam and cancel afterwards; that costs less than failing, waiting 14 days, and paying the full fee again.

Several vendors have standing ways to pay less, worth knowing before you register:

• AWS: the 2026 prepaid Professional/Specialty exam voucher includes one free retake, but under the promotion terms (in Mandarin) the voucher had to be bought by 2026-06-30 and the first attempt taken by 12-31. If you already hold one, a failed AIP-C01 attempt does not cost another $300.
• Google: renewal comes with a 50% code, and Google Cloud customers can get free exam-prep training through Get Certified; both are covered in the Google section above.
• Microsoft: the beta period of a new certification usually comes with a public code giving the first 300 candidates 80% off, published in that certification's announcement. The AI-500 offer ended 2026-08-05.
• Databricks: completing a listed learning pathway during a Virtual Learning Festival earns a 50%-off certification voucher (worth $100) and a 20% Academy Labs coupon. The most recent one ran 2026-03-16 to 04-03.
• NVIDIA: on-site proctored certification exams at GTC are included with the conference pass.

Ask your employer first. Enterprise programs and team exam vouchers are usually worth more than any discount an individual can find, so check for an internal allocation before registering.

Three Things That Are Not Individual AI Certifications

These three show up on recommendation lists, but none is an AI certification you can sit: one does not exist, one is not a certification, and one is not awarded to individuals.

"Google Cloud GenAI Engineer" does not exist. The AI-related entries in Google's official certification index are Generative AI Leader, Professional ML Engineer, and the new Professional Agentic Architect; nothing carries this name.

Free course certificates from OpenAI and Anthropic are not certifications. Both vendors say so outright: the OpenAI Academy help page states completion certificates "are not OpenAI Certifications, do not represent a formal OpenAI credential," and the Claude Academy FAQ answers the "is this the same as certification" question with "No — they're two different things." The courses are worth taking, but what they attest to is course completion, not assessed capability — they do not belong in a field asking for a professional AI/ML certification. OpenAI's actual proctored credential remains in employer pilots, with no public registration, price, or date.

CNCF's "Certified Kubernetes AI Conformance" is a real certification, just not one awarded to people. It is a conformance program for platforms and vendors, certifying that a Kubernetes distribution can reliably run AI workloads, awarded on a vendor-submitted self-assessment. CNCF's individual certification list (CKA, CKAD, CKS, KCNA, CNPA, CNPE, and so on) currently contains no AI-specific credential at all.

Three Things to Check Before You Register

Study material freshness matters more than the certification itself. PMLE is the clearest case: same name, same price, same-looking landing page, but old material walks you into a wall. Ten minutes with the official exam guide comparing service names before you register beats another hundred practice questions.

Specs are often not on the certification landing page. The fees, item counts, validity, and domain weights for the four Claude exams are in their exam guide PDFs, and NVIDIA's full blueprints are in PDF study guides too. When checking specs, follow the path from the certification page to the exam guide or blueprint download.

Check the update date on official pages too. Community Q&A and older announcements often lag the certification page by months — Microsoft Q&A still hosts answers saying AI-103 is in beta, long after the certification page stopped saying so. Go by the certification page and the study guide.

Overall

Match the cloud first, then the certification. None of these vendors' credentials transfer; picking the wrong ecosystem adds almost nothing to a résumé. The one exception is CertNexus CAIP — ISO/IEC 17024 accredited and vendor-neutral, at the cost of proving nothing about any specific platform.

Put validity into the total cost. AWS is 3 years, Google 2, NVIDIA and Databricks 2. The AWS AI trio renews by exam only, but a 50% voucher applies and passing a higher exam renews the lower ones; PMLE also requires a retake today, but comes with a 50% renewal code, and continuing-education renewal already covers four other Google certifications with more promised. The free online renewal assessments from Microsoft and Claude are the cheapest of the group.

Certifications have a shorter half-life than people assume. Microsoft turned over its entire AI track, PMLE renamed every product in its guide, and Claude went from one exam to four — all within the first half of 2026. Assume any certification recommendation older than three months has at least one dead line in it, this one included; open the vendor page before you pay.

Taiwan's iPAS intermediate remains good value — NT$1,000 for both subjects, a national credential with real weight in local hiring and government procurement — but budget for the 5-year renewal.

The governance line gates on something else entirely. AIGP, AAISM, and AAIA are not "harder AI certifications"; they are add-ons that presuppose a compliance or security identity. Without CISM / CISSP / CISA you cannot sit the ISACA pair at all, regardless of whether you could pass.

Changelog

• 2026-10-12: Added "Beyond the Exam Fee: Course Costs and Ways to Pay Less"; the Google section now includes Professional Agentic Architect, opening for registration in November, and PMLE chapter 6 is given as about 13%; iPAS registration for the second session has closed; added the AWS retake rule, the Claude training gate, and iPAS question sets; rewrote "Three Things That Are Not Individual AI Certifications" and "Three Things to Check Before You Register".
• 2026-10-12: The series entry now links to the series page and the preparation-method article. The AWS section follows the current state: MLA-C01 in English ended 9/28 and the MLA-C02 English beta is what is delivered now, with a link to the MLA-C02 preparation path; "the only one offering Traditional Chinese" now reads as the only one of AWS's three.
• 2026-10-08 (recheck): the AI-500 pages no longer carry a beta label and the wording is updated; added that AB-100 now requires an associate certification; added Google Professional Agentic Architect; the old URL for the Google PMLE exam guide is dead and now points to the official PDF.
• 2026-10-08: Corrected two errors in the GH-300 paragraph. It said the official page "publishes no figure"; the price loads dynamically by country and is $99 in the US and $50 in Taiwan. It said Microsoft's free renewal assessment "extends to" GH-300; GitHub certifications are valid for 2 years and the renewal process is still in transition. Also added links to five new posts in the series (AI-901, AI-200, AI-300, GH-300, GH-600).
• 2026-08-19: Added links to the 24-post AI Certification Prep series. Fifty files linked to this article; it linked back to one. The "which one" table now leads into the series and the four per-vendor choosers, and each vendor section ends with its preparation paths. No content or conclusions changed.
• 2026-08-18: Substantially expanded coverage. NVIDIA goes from a single exam to the full twelve-exam catalog (including Agentic AI NCP-AAI and multimodal NCA-GENM), correcting the earlier claim that only Microsoft offers an expert-level agent credential; added "Other Ecosystems" (Snowflake GES-C02, Oracle's new Agentic AI track, IBM watsonx C1000-185, Salesforce Agentforce Specialist AI-201) and GitHub Copilot GH-300; added "Governance, Audit, Security" (IAPP AIGP, ISACA AAISM / AAIA, CertNexus CAIP); added iPAS beginner-level specs plus the intermediate certificate split, subject-score carry-over, and exemption rules, along with TQC and TQC+; added AI-500's hard prerequisite (AI-103 must be earned first, quoted from the official Certification prerequisites section) along with its beta and English-only status; added the MLA-C01 update notice (English retires 2026-09-28, C02 registration opens September 1 with no published guide) and the exam-language details (AIF-C01 is the only one offered in Traditional Chinese); corrected the AWS renewal section — the article said renewal meant retaking with no continuing-education option, whereas the official recertification page lists both renew and maintain paths (maintain limited to SAA, Developer, CloudOps, SAP, DOP), cross-certification renewal across the AI trio, and a 50% voucher on every renewal; added a "commonly mis-listed" section covering Google Cloud GenAI Engineer (does not exist), OpenAI (free, globally available course certificates that are explicitly not certifications, with the formal credential still in employer pilots), and CNCF Kubernetes AI Conformance (awarded to platforms, not people), kept as three distinct errors rather than one shared reason. Per the "professional AI/ML certifications only" scope, free course-completion certificates from OpenAI Academy and Claude Academy are not listed as credentials; a single note with official sources remains in the mis-listed section. A significant error in the Claude section was corrected: the article claimed Anthropic published no prices or validity, when in fact each exam guide PDF states the fee ($99 / $125 / $125 / $175), item count, 12-month validity, and domain weights; retake rules were added, and the "what a vendor doesn't publish isn't known" takeaway was rewritten as "not finding it is not the same as it not existing." Retired exams were dropped from the article so that everything listed is currently registrable. Three existing facts were re-verified: the Google certification index (2 foundational, 3 associate, 9 professional) confirms Generative AI Leader and PMLE are still the only AI credentials, so that conclusion stands; the generative-ai-engineer evidence was corrected — that URL is a soft 404 (404 page body, HTTP status 200), so the original "returns HTTP 404" was inaccurate; and two Google renewal changes were added, namely the 50% renewal code issued at first certification and the continuing-education path currently limited to CDL, ACE, PCA, and PDE, leaving PMLE exam-only.

References

AWS

• AWS Certified Generative AI Developer – Professional (AIP-C01)
• AWS Certification: After Testing policy (retake wait and result timing)
• AWS Skill Builder subscription plans and pricing
• AWS voucher promotion terms and conditions (Professional/Specialty with free retake)
• Google Cloud Certification: Vouchers & Discounts (50% renewal code)
• Databricks Virtual Learning Festival FAQ (50% voucher rules)
• NVIDIA Training (GTC on-site training and certification exams)
• AWS Certified Machine Learning Engineer – Associate (MLA-C02)
• AWS Certified AI Practitioner (AIF-C01)
• AWS Recertification policy (three-year validity)

Google Cloud

• Professional ML Engineer official exam guide
• Professional ML Engineer certification page
• Generative AI Leader certification page
• Google Cloud certification index (includes Get Certified program)
• Google Cloud Exam Terms & Conditions (validity and renewal)
• Google Cloud Certification Renewal (renewal paths, 50% code, continuing-education scope)
• Welcome to Google Cloud Next '26 (Gemini Enterprise Agent Platform launch)
• GEAR program and AI learning paths

Microsoft

• Azure AI Apps and Agents Developer Associate (AI-103)
• Multi-Agent AI Solutions Expert (AI-500)
• Agentic AI Business Solutions Architect (AB-100)

NVIDIA / Databricks / Anthropic

• NVIDIA certification catalog (all twelve exams and prices)
• NVIDIA NCA-GENL official page
• NVIDIA NCP-AAI (Agentic AI Professional)
• NVIDIA NCA-GENM (Generative AI Multimodal Associate)
• Databricks Certified Generative AI Engineer Associate
• Anthropic: Four role-based Claude certifications (2026-07-23)
• Pearson VUE — Claude Certification Program (retake rules)
• Anthropic Partner Academy — the four certification pages and exam guide downloads
• Claude Academy FAQ (free completion badges vs. proctored certification)

Other ecosystems

• Snowflake SnowPro certification index (GES-C02 and exam pricing)
• Oracle: OCI Certification Learning Paths and Exams, 2026 updates
• IBM Certified watsonx Generative AI Engineer – Associate (C1000-185)
• Salesforce Certified Agentforce Specialist (AI-201)
• GitHub Copilot certification (GH-300) exam details
• Exam GH-300 study guide (Microsoft Learn)

Governance / audit / security

• IAPP AIGP exam page (pricing and maintenance)
• ISACA AAISM certification page
• ISACA AAIA certification page
• CertNexus CAIP (AIP-210) certification page
• CAIP AIP-210 exam blueprint (domain weights and recertification)

Commonly mis-listed

• OpenAI Academy courses (free, global, and explicitly not certifications)
• OpenAI: Launching our first OpenAI Certifications courses
• CNCF launches the Kubernetes AI Conformance Program (platform-level, not individual)
• CNCF individual certification list

iPAS and Taiwan-local

• iPAS AI Application Planner exam information (in Mandarin)
• 2026 iPAS AI Application Planner examination handbook (in Mandarin)
• Notice on intermediate-level code question weighting (in Mandarin)
• iPAS exam registration instructions (in Mandarin)
• 2026 iPAS handbook, April revision (certificate split and exemption rules) (in Mandarin)
• TQC AI Applications and Technology (in Mandarin)
• TQC+ Artificial Intelligence: Machine Learning Python 3 (PML3) (in Mandarin)
• 104: 2026 AI certification guide (in Mandarin)

Related on this site

• Complete Guide to the Claude Certified Architect Foundations Exam
• Which AI courses to take in 2026 (in Mandarin)